Benchmark Keras Model Using TensforFlow Benchmark
I am trying to compare the inference phase performance of my Keras model using TensorFlow support. I thought the Tensorflow Benchmark tool was the right way.
I was able to build and run the example on desktop with tensorflow_inception_graph.pb
and everything works fine.
I can't figure out how to store the Keras model as the correct model .pb
. I can get TensorFlow plot from Keras model like this:
import keras.backend as K
K.set_learning_phase(0)
trained_model = function_that_returns_compiled_model()
sess = K.get_session()
sess.graph # This works
# Get the input tensor name for TF Benchmark
trained_model.input
> <tf.Tensor 'input_1:0' shape=(?, 360, 480, 3) dtype=float32>
# Get the output tensor name for TF Benchmark
trained_model.output
> <tf.Tensor 'reshape_2/Reshape:0' shape=(?, 360, 480, 12) dtype=float32>
I am trying to save the model in different ways.
import tensorflow as tf
from tensorflow.contrib.session_bundle import exporter
model = trained_model
export_path = "path/to/folder" # where to save the exported graph
export_version = 1 # version number (integer)
saver = tf.train.Saver(sharded=True)
model_exporter = exporter.Exporter(saver)
signature = exporter.classification_signature(input_tensor=model.input, scores_tensor=model.output)
model_exporter.init(sess.graph.as_graph_def(), default_graph_signature=signature)
model_exporter.export(export_path, tf.constant(export_version), sess)
Which creates a folder with some files, I don't know what to do.
Now I run the Benchmark tool with something like this
bazel-bin/tensorflow/tools/benchmark/benchmark_model \
--graph=tensorflow/tools/benchmark/what_file.pb \
--input_layer="input_1:0" \
--input_layer_shape="1,360,480,3" \
--input_layer_type="float" \
--output_layer="reshape_2/Reshape:0"
But no matter what file I try to use as what_file.pb
, I getError during inference: Invalid argument: Session was not created with a graph before Run()!
source to share
So I got this to work. You just need to convert all the variables in the tensorflow graph to constants and then save the graph definition.
Here's a small example:
import tensorflow as tf
from keras import backend as K
from tensorflow.python.framework import graph_util
K.set_learning_phase(0)
model = function_that_returns_your_keras_model()
sess = K.get_session()
output_node_name = "my_output_node" # Name of your output node
with sess as sess:
init_op = tf.global_variables_initializer()
sess.run(init_op)
graph_def = sess.graph.as_graph_def()
output_graph_def = graph_util.convert_variables_to_constants(
sess,
sess.graph.as_graph_def(),
output_node_name.split(","))
tf.train.write_graph(output_graph_def,
logdir="my_dir",
name="my_model.pb",
as_text=False)
Now just call the TensorFlow Benchmark tool my_model.pb
as a graph.
source to share
You are saving the parameters of this model, not the definition of the graph; to save this usage tf.get_default_graph().as_graph_def().SerializeToString()
and then save it to a file.
This suggests that I don't think the reference tool will work as it doesn't have the ability to initialize the variables your model depends on.
source to share